Suppose I'm running a website on a VPS or dedicated server running on LAMP. If my page generation time is ~20ms on average, does this mean I can handle ~50 requests per second on one cpu core? Do I have to take user bandwidth into account? (E.g. if it takes 500ms to send the data, can I only count on handling 2 requests per second, or is it limited only by the server's outgoing bandwidth?) I have no actual data to share, since I'm working on a not yet launched website.
You cannot calculate how much your server will handle based on the response time of your website. Every server has different specification such as the processor and the memory. You can determine if you server will be able to handle the data by determining how much traffic you will be receiving to your website on average a month. Also what kind of website is it gonna be: file sharing website, a simple static page site, etc... will impact the performance of the server.
Doesn't that seem a bit cyclical? In order to determine how much traffic I can handle I have to determine how much traffic I get? Also, I expect most traffic to go through the 99% static main page.
If you only need traffic info ask your hsoting provider what is the uplink speed. How much bandwidth you are allowed . You can calculate based on that.